16 ms·

Gah. Where do we even begin with this article's problems? 1. The studies were done with college students (a very unrepresentative, and relatively immature grou

by zeteo 16y ago

Gah. Where do we even begin with this article's problems? 1. The studies were done with college students (a very unrepresentative, and relatively immature group), in a non-organization (basically they were linked only by a possible monetary reward at the end) and with some very trivial incentive in play. If this has "implications for business work groups, volunteer organizations, non-profit projects, military units, and environmental efforts", please let us know how those implications are arrived at! 2. The description of the experiment is unclear and self-contradictory: "...pools of points that they could keep or give up for an immediate reward of meal service vouchers. Participants were also told that giving up points would improve the group's chance of receiving a monetary reward." So if you "give up" points you _both_ get free meals _and_ you improve the group's welfare?! That doesn't make any sense. 3. What's the deal with the rules - what rules were the "do-gooders" breaking? They're only mentioned in passing but they seem essential for the meaning of the experiment.

yummyfajitas 16y ago
Regarding point 1), it's even worse than that. College students live in an environment where "do-gooder" behavior objectively harms them. If I slack off and you study hard, the professor is comparing my 40/100 to your 85/100. This results in me getting a lower letter grade.
- zephyrfalcon 16y ago
  To be honest, I never understood this letter grading system used in the US (and maybe elsewhere)... and if you already have scores (like 40/100 vs 85/100), why even bother with another system?
  - yummyfajitas 16y ago
    A = 1 standard dev above the mean. B = mean. C = 1 std dev below the mean, but still performed remotely adequately. D = "If your major is theater and the class is math, I won't stop you from graduating."
    - philwelch 16y ago
      Traditionally (and as is still the case at some schools, according to a friend of mine now at Georgia Tech), the scale was: A = Exceptionally above the mean. B = 1 stdev above the mean. C = mean. D = 1 stdev below the mean. F = More than 1 stdev below the mean. At most schools, however, grade inflation has changed the scale upwards to what you said.
      - po 16y ago
        It can even vary within schools. At my university, the classes in my major (Chemistry) were graded as you describe but other classes I took (like Psychology) used the inflated scale.
      - ams6110 16y ago
        Interestingly, in many of my middle- and high-school classes, grades were not based on standard deviations from the mean of actual class performance; they were a fixed scale: 90% or better was an "A", 80 - 89% was a B, 70 - 79% was a "C", etc. So you really could have the entire class be "better than average" (by the more usual method of computing a "C" grade) if the teacher made an assignment too easy. My son's school still does this, though they've slid the scale up a bit, you need a 93% or better now to get an "A", and they've added a safety net: if you turn in any kind of reasonable attempt at all work assigned you will not get lower than a D- in the class.
        philwelch 16y ago
        There's no requirement that you curve the scores to any real distribution after the fact--you can design your tests with the intention that the average student gets 70-79% of the questions correct and students in other quintiles get higher or lower scores, or with the intention that the average student gets 80-89% of the questions correct, or with some other mix of intentions and related test scores. With some experience and fiddling I bet you could get a grading system like you describe which fell along a standard distribution.
  - Daniel_Newby 16y ago
    To account for the unpredictable variability of test difficulty.
Confusion 16y ago
4. The situation is completely artificial, to the extent that experimental partners are simulated and behave randomly. The conclusion could as well be that people don't like erratic behavior, even if the unpredictable behavior is do-goody. I can completely empathize with that: don't we all want to be able to depend on a colleague? To be sure that something that gets promised gets done and gets done well? Someone that behaves erratically is simply less trustworthy.

by zeteo 16y ago

yummyfajitas 16y ago

Regarding point 1), it's even worse than that. College students live in an environment where "do-gooder" behavior objectively harms them. If I slack off and you study hard, the professor is comparing my 40/100 to your 85/100. This results in me getting a lower letter grade.

zephyrfalcon 16y ago
To be honest, I never understood this letter grading system used in the US (and maybe elsewhere)... and if you already have scores (like 40/100 vs 85/100), why even bother with another system?
- yummyfajitas 16y ago
  A = 1 standard dev above the mean. B = mean. C = 1 std dev below the mean, but still performed remotely adequately. D = "If your major is theater and the class is math, I won't stop you from graduating."
  - philwelch 16y ago
    Traditionally (and as is still the case at some schools, according to a friend of mine now at Georgia Tech), the scale was: A = Exceptionally above the mean. B = 1 stdev above the mean. C = mean. D = 1 stdev below the mean. F = More than 1 stdev below the mean. At most schools, however, grade inflation has changed the scale upwards to what you said.
    - po 16y ago
      It can even vary within schools. At my university, the classes in my major (Chemistry) were graded as you describe but other classes I took (like Psychology) used the inflated scale.
    - ams6110 16y ago
      Interestingly, in many of my middle- and high-school classes, grades were not based on standard deviations from the mean of actual class performance; they were a fixed scale: 90% or better was an "A", 80 - 89% was a B, 70 - 79% was a "C", etc. So you really could have the entire class be "better than average" (by the more usual method of computing a "C" grade) if the teacher made an assignment too easy. My son's school still does this, though they've slid the scale up a bit, you need a 93% or better now to get an "A", and they've added a safety net: if you turn in any kind of reasonable attempt at all work assigned you will not get lower than a D- in the class.
      - philwelch 16y ago
        There's no requirement that you curve the scores to any real distribution after the fact--you can design your tests with the intention that the average student gets 70-79% of the questions correct and students in other quintiles get higher or lower scores, or with the intention that the average student gets 80-89% of the questions correct, or with some other mix of intentions and related test scores. With some experience and fiddling I bet you could get a grading system like you describe which fell along a standard distribution.
- Daniel_Newby 16y ago
  To account for the unpredictable variability of test difficulty.

Confusion 16y ago

4. The situation is completely artificial, to the extent that experimental partners are simulated and behave randomly. The conclusion could as well be that people don't like erratic behavior, even if the unpredictable behavior is do-goody. I can completely empathize with that: don't we all want to be able to depend on a colleague? To be sure that something that gets promised gets done and gets done well? Someone that behaves erratically is simply less trustworthy.